Just thinking out loud, perhaps start with calculating perceptual hashes [1] of frames in both videos and use the inter frame hash distances of videos as finger

Just thinking out loud, perhaps start with calculating perceptual hashes [1] of frames in both videos and use the inter frame hash distances of videos as fingerprints. As I understand it, perceptual hashing is robust against both resolution and color variations. [1]: http://www.hackerfactor.com/blog/?/archives/432-Looks-Like-It.html http://www.hackerfactor.com/blog/?/archives/432-Looks-Like-I...